Annual Journal of Hydraulic Engineering, JSCE, Vol.54, 2010, February
DIFFERENT SKILLS OF FIVE GCMs AND THEIR
IMPACTS ON AQUIFER THERMAL REGIMES
Luminda GUNAWARDHANA1 and So KAZAMA2
1Member of JSCE, M.Sc., Graduate School of Environmental Studies, Tohoku University (Aoba 20, Sendai,
980-8579. Japan)
2Member of JSCE, Dr. of Eng., Graduate School of Environmental Studies, Tohoku University (Aoba 20, Sendai,
980-8579. Japan).
For the purposes of estimating local changes in aquifer thermal regimes, induced by global climate
change, we analyzed different performances of six general circulation models (GCMs) in the Sendai plain.
Transfer function method was used to downscale the GCMs output to the local scale. GCM scenarios
were evaluated based on their match with the 20th century observations and the magnitude of climatic
parameter change in 21st century. Among the considered GCMs; HADCM3, MIROC and MRI models
produced climate variations in a wide and consistent range (2.0 to 4.7 °C warming and -11 to 345 mm
precipitation change) suggesting their applicability for climate change studies in the Sendai plain. When
accounting the effects of the entire model scenarios, groundwater recharge would vary in range of 50-182
mm/year and aquifer temperature may change in 1.2-3.9 °C range in year 2080 which, according to IPCC
AR4, may have critical impact on the ecological balance of the Sendai plain.
Key Words : Climate downscaling, transfer function, groundwater temperature, Sendai plain
1. INTRODUCTION
The unavoidable climate change impacts are
often expressed through hydrological responses
such as flooding and groundwater quantity
depletion, but occasionally evaluated in terms of
groundwater quality1),2). Groundwater temperature is
one of the primary parameter regulating the
ecological balance of the groundwater dominated
ecosystems3), and thus highly vulnerable to the
changing climate. Assessing these impacts demand
future scenarios of global climate behavior. The
Fourth Assessment Report of Intergovernmental
Panel on Climate Change (IPCC AR4)4) provides a
platform for climate change studies with a set of
widely recognized Global Climate Models (GCMs).
However, reliable assessment of anticipated climate
change impacts eventually requires fine resolution
projection (~10 km2) of climate parameters at the
local scale, which cannot be resolved by current
GCMs. Model predictions with such high
resolutions are computationally demanding and not
likely to become widely available in the near future.
For example, HADCM3 model has very coarse
- 67 -
resolution, which in spatial scale, approximately
equal to 90465 km2 of the grid box containing the
Sendai plain. Even though, some models like
MIROC3_2-HI and CCSM3 recently produce model
predictions with comparatively higher resolution
(12185 km2 and 19070 km2 resolutions reference to
the Sendai plain, respectively), they are still far
behind to be used for site specific assessments. As a
result, GCMs cannot explicitly account for the
physical-geographic characteristics of the fine scale
structure (e.g. inland water bodies, coastal lines,
mountain ranges, and land use) that significantly
govern the local climate. However, GCMs show a
good performance in simulating large scale
circulation and climatic features that affect regional
climates5). Therefore, statistical downscaling
technique has long been used as an intermediate step
between coarse resolution GCM output and fine
scale climate variables1),2),5).
The ability to incorporate specific local
information and fact that the method is
computationally inexpensive, make statistical
downscaling technique reliable and easy to apply
over different GCMs. However, in addition to the
differences in grid resolutions, each GCM scenario
incorporates different model structures that produce
different output. Therefore it is needed to use
several GCMs and scenarios for a proper impact
assessment rather than relying on single forecast5).
To date, there has not been sufficient research to
examine the different behaviors of various GCMs. A
detail analysis of several GCMs and scenarios may
facilitate many site specific impact assessment
studies to select few significant model scenarios in a
range of potential climate change in future.
Therefore, our objectives are two-fold. First we will
present some different performances of six GCMs
for predicting the future climate change at the local
scale in the Sendai plain. For the second objective,
downscaled results from GCMs, which provide 18
simulations for 21st century, will be used to evaluate
the range of aquifer temperature change by year
2080. These results with the potential range of
impact on groundwater temperature may guide the
decision makes on making resilient decisions for
climate impact mitigation measures.
2. METHODOLOGY
(1) Study area and data collections
The area between the Nanakita and Natori rivers
in the Sendai plain was selected. The Sendai plain is
an alluvial formation and serves as the main aquifer
of the catchment. The maximum depth of the
aquifer ranges between 60−80 m and the
permeability of the soil below that is significantly
(approximately 104 times) less than the permeability
of the main aquifer6). Similar to Fukushima and
Yamagata cities near by study area, surface air
temperature (SAT) records in Sendai meteorological
station indicate no significant trend until the middle
of 20th century. Since 1947 until 2007, SAT shows
significant increasing trend in the region (e.g. 2.0,
1.8, and 1.7 °C/century in Sendai, Yamagata, and
Fukushima, respectively). In contrast to temperature,
annual total precipitation in the Sendai plain shows
no strong trend over the last 80 years.
There are five water level observations stations
located within the area (Fig. 1). Among them, W1,
W2, W4 and W5 have three sub-wells (SWs) each
directed to different aquifer depths (e.g. 7, 26, 60 m
at W1). Groundwater temperature was measured at
W1, W2, W3 and W4 at 1 m intervals. Groundwater
temperatures presented in Water Environmental
Map No. 1 were used for W5. Groundwater levels
measured by the Sendai city office were also taken.
Some significant differences in land use types exist
at the local scale surrounding the wells. Wells W2,
W3 and W4 are located in paddy field areas and
have similar surface characteristics and land use
- 68 -
Fig.1 Study area.
histories. Well W1 is located in a residential area
and well W5 is located at the city center, which is
more urbanized than all other well locations. All
well locations are situated within seven kilometers
of the city center and the Sendai meteorological
station. In all wells, there is clear evidence of a
temperature profile inversion from the general
geothermal gradient. The magnitude of ground
surface warming, which was calculated as the
difference between the observed temperature depth
(T-D) profile and the extrapolated steady state linear
curve to the ground surface, ranges 0.9−1.3 ºC.
(2) Analytical model for sub-surface heat flow
Horizontal water flow in a semi−confining
aquifer can significantly affect the vertical
temperature distribution in the basin. Monthly
averaged water level records in 2007 produced
vertical hydraulic gradient in 0.077~0.169 m/m and
horizontal hydraulic gradient in 0.0008~0.0014 m/m
ranges. According to that, averaged vertical
hydraulic gradient is notably higher than the
averaged horizontal hydraulic gradient, which may
confirm the appropriateness of the one-dimensional
heat transport model for the study area.
Temperature distribution in one-dimensional
homogeneous porous media with constant
incompressible fluid flow can be described as
α (∂ 2T ∂z 2 ) − β(∂T ∂z ) = ∂T ∂t
(1)
where T is temperature, z is depth from the ground
surface, t is time, α (= k/cρ) is the thermal diffusivity
of the aquifer and β = vc0ρ0/cρ, where v is the
vertical groundwater flux (positive is downward),
c0ρ0 is the heat capacity of the water and cρ is the
heat capacity of the porous medium. Assuming that
the heat capacity is constant and that the
groundwater and aquifer are in thermal equilibrium,
the initial boundary conditions for a linear increase
in ground surface temperature can be written as
T( z ,0 ) = T0 + az
(2)
(3)
where T0 is the ground surface temperature at t = 0,
a is the general thermal gradient, and b is the rate of
surface warming. Under the above initial boundary
conditions, Carslaw and Jaeger7) obtained an
analytical solution for the temperature distribution
as a function of depth and time:
T = T0 + a ( z − β t ) + {(b + β a ) / 2β} × [( z + β t ) exp(β z / α )
erfc{(z + β t)/2(αt) 1/2 } + (β t − z )erfc{(z - β t)/2(αt) 1/2 }] (4)
3. RESULTS AND DISCUSSIONS
(1) Model calibration and verification
The undisturbed depths of the aquifer (e.g.,
below 60 m at W5) exhibit an approximately linear
T−D profile. Extrapolation of this linear portion to
the ground surface yields the intercept temperature
(T0). If thermal conductivity is assumed to be
uniform over the representative depth interval, the
general geothermal gradient (a) can be estimated
from the gradient of the undisturbed portion of the
T−D profile. Therefore, the parameters a and T0 in
Eq. (4) can be estimated with reasonable accuracy
(±0.002ºC /m, ± 0.1ºC, respectively) from the
estimated T−D profile in each observation well. A
series of synthetic T−D profiles over different
values of b, t and v were then formulated and
compared with the observed T−D profile in the
Sendai plain for a preliminary approximation of the
parameters in Eq. (4). The parameter t was
calibrated to be 60 years (1947-2007), which
reasonably agrees with the starting time of SAT
warming in the Sendai plain. The parameter b was
calibrated for each observation well, which spans in
1.5~2.1 °C/100years. It shows good agreement with
the land-use type and depth of departure from the
steady state temperature in all well locations.
Moreover, observed linear trend of SAT warming in
the Sendai plain (2.0 °C/100years) lays in calibrated
range suggesting a fine match with the local climatic
conditions. Considering the potential uncertainties
of groundwater recharge estimation, Gunawardhana
and kazama2) estimated the groundwater recharge
rate by three other techniques (the water balance
method, the water level fluctuation method and
Darcy’s method) that produced groundwater
recharge rates over a consistent range of 105-210
mm/year in the Sendai plain. The calibrated
groundwater recharges by the T-D profile method in
different well locations respect to various land-use
types range 105-215 mm/year, showing a good
match with the previous studies. Simulated results
with constrained parameters are shown in Fig. 2,
which shows good agreement with the observed
T-D profiles.
- 69 -
(2) Assessment of climate model and scenarios
To assess the performance of the different
models in capturing local climate in the Sendai plain,
we compare the time series of each model against
the observations during 1927-1999. Fig. 3a) shows
the linear trends of temperature of different GCM
scenarios with a comparison of observed trend over
73 years. Many researches from IPCC AR44)
suggested that global averaged warming in 20th
century ranges from 0.5-0.7 °C/century. Of the six
GCMs analyzed, four of them (HADCM3, MRI,
ECHAM5, and CSIRO) simulate warming trend in
20th century (black color boxes in Fig. 3a) for the
grid box containing the Sendai plain. CCSM3 model
simulates very little trend while MIROC exhibits
negative trend. Even though the temperature trend is
less dependent on grid resolution, GCMs usually
unable to capture the local effects of urban heat
island of the size of Sendai city. This may explain
the mark of significantly higher warming rate in the
Sendai plain than the simulations from six GCMs
(Observed warming rate is over two times higher
than the HADCM3 model, which simulates the
highest warming rate among six GCMs). According
to the warming trends in the future, HADCM3
model predicts the highest warming rate in the end
of 21st century while MRI model shows the smallest
trend, which is almost similar to the present trend of
warming rate in the Sendai plain. However, in
accordance with the IPCC AR44), magnitude of
global warming in later part of the 21st century
expects to be significantly higher than the 20th
century. Thus, it is reasonable to assume that the
simulations from HADCM3 may not over predict
the future but a possibility and therefore, can be
used as a potential impact predictor for the future
studies in the Sendai plain.
Even though there are some significant
geographical variations of climate change,
particularly with respect to precipitation, globally
mean precipitation is projected to increase in future.
According to IPCC AR44), high latitudes countries
(e.g. Japan) likely experience high precipitation
events. As depicts in Fig. 3b, five GCMs predict
0
Temperature ( C)
0
13.6
-10
13.8
14
14.2
14.4
14.6
14.8
15
15.2
-20
Depth (m)
T(0 ,t ) = T0 + bt ,
-30
-40
-50
-60
-70
W1-Observed
W1-Simulated
W2-Observed
W2-Simulated
W3-Observed
W3-Simulated
W4-Observed
W4-Simulated
W5-Observed
W5-Simulated
Fig. 2 Model verification with observed T-D profiles
potential of increasing precipitation in the 21st
century while only MRI model predicts decreasing
trend. HADCM3, MIROC and CCSM3 models
simulate significant precipitation change, which
predict over 300 mm of annual total precipitations
rise in 21st century relative to the observations in
20th century. When comparing GCM simulations
with observed precipitation in same time horizon
(1927-1999 in black color in Fig. 3b), CSIRO and
HADCM3 exhibit comparatively smaller biases
(55mm and 149mm, respectively) while CCSM3
shows the strongest wet bias (601 mm). In contrast,
MRI model shows dry bias, but similar to the
decreasing pattern of precipitation in future.
a)
b)
4.5
4.0
From 1927-1999
3.5
From 1927-2099
1000
Averaged (from 1927 to 1999) Annual total precipitation difference of GCM
simulations relative to observations in same time period
800
Averaged (from 2000 to 2999) Annual total precipitation difference of GCM
simulations relative to observations in 1927-1999 time period
o
200
CCSM3-B1
CCSM3-A2
CCSM3-A1B
MIROC-B1
CCSM3-20C3M
MIROC-A2
MIROC-A1B
HADCM3-B1
MIROC-20C3M
HADCM3-A2
HADCM3-A1B
ECHAM5-B1
ECHAM5-A2
HADCM3-20C3M
-400
ECHAM5-A1B
-200
CSIRO-B1
0
ECHAM5-20C3M
MIROC-B1
MIROC-A2
MIROC-A1B
CCSM3-B1
MIROC-20C3M
CCSM3-A2
CCSM3-A1B
CSIRO-B1
CCSM3-20C3M
CSIRO-A2
CSIRO-A1B
ECHAM5-B1
CSIRO-20C3M
ECHAM5-A2
ECHAM5-A1B
MRI-B1
ECHAM5-20C3M
MRI-A2
MRI-A1B
MRI-20C3M
HADCM3-B1
HADCM3-A2
HADCM3-A1B
Observed
-0.5
HADCM3-20C3M
0.0
CSIRO-A2
0.5
400
CSIRO-A1B
1.0
CSIRO-20C3M
1.5
600
MRI-A2
2.0
MRI-20C3M
2.5
Precipitation change (mm)
3.0
MRI-A1B
MRI-B1
Linear trend of temperature change ( C/100 years)
(3) Spatial downscaling of GCMs output
In order to estimate the global climate change
impacts at the Sendai plain, we examined time
series of temperature and precipitation for the
20C3M, A2, A1B and B2 scenarios of all 6 models,
which produce 6 and 18 time series for each
climatic parameter for the 20th and 21st centuries,
respectively. Model results containing the target
station in the Sendai meteorological station (38.26
o
N and 140.9 oE) were spatially downscaled using
the transfer function method (more details in
Gunawardhana and Kazama1),2)). Transfer function
method builds a regression relationship between
GCM output (predictors) and local climate variable
(predictand). Among the parameters of concern, sea
level pressure has significant influence to govern the
local precipitation. However, many studies found
that the direct use of GCM precipitation as the sole
predictor produce better local precipitation than
introducing multiple predictors5). Similarly, GCM
scale surface air temperature is a robust predictor for
regional temperature. Therefore, we considered
large scale climatic parameter as the only predictor
governing the representative climate variable at the
local scale. Projected GCMs data of 1967-2006
were used as the control to developed transfer
functions in conjunction with measured data of the
same period. Those transfer functions were then
used to downscale the 1947-1966 GCM data to the
Sendai plain and later verified with the observed
climatic parameter. For climate prediction, those
transfer functions were further used to downscale
2060-2099 GCM precipitation and temperature. Fig.
4 shows the comparisons of cumulative probability
distributions of temperature and precipitation.
According to Fig. 4a, the highest warming was
projected by MIROC-A2 scenario (about 5.0 °C in
January and 6.6 °C in July), which cause average of
4.7 °C warming in 2060-2099 time periods relative
to 1967-2006. For the same model scenario,
precipitation is projected to increase by 18% (217
mm) relative to observed precipitation during
1967-2006 time periods in the Sendai plain.
Model bias in GCM simulations weakens the
magnitude of GCM predictions at the local scale.
Therefore, when considering several GCM models,
scenario that predicts the highest warming rate at the
GCM grid scale will not necessarily predict the
utmost impact in the local scale. According to Fig.
3a, HADCM3 model predicts the highest warming
trend at the GCM grid scale, but at the local scale it
was dropped to the second by MIROC model, which
predicts 0.8 °C warming than the HADCM3 in
2080. This change was occurred due to bias
difference between two models relative to the
observed trend. Therefore, the criteria of selecting
the GCMs to produce extreme impacts must include
not only the magnitude of climatic parameter
change in the future (e.g. annual precipitation
change during 2000-2099 relative to 1927-1999 in
this study), but also the bias of GCMs relative to the
observations at the local scale. Therefore, two
indexes with accounting the model bias and the
magnitude of climatic parameter change in future
were considered for assessing the performance of
different GCMs in the local climate predictions. As
example, for the temperature, relative difference
Fig. 3 Comparisons of observed climatic parameters with the different GCM scenarios; a) linear trend of temperature change and b)
averaged annual total precipitation change relative to the observations during 1927-1999.
- 70 -
projected 18% increase in precipitation, due to
higher degree of evapotranspiration resulting from a
4.7 °C increase in surface air temperature. However,
due to its slow rate of warming (the second lowest
among 5 GCMs) and moderate precipitation rise,
ECHAM5 model predicts 11-35% of recharge
increment in 2080 for all model scenarios. Moreover,
combined effects of groundwater recharge variation
(as β) and ground surface temperature change (as b)
were applied in Eq. (4) to estimate the potential
aquifer temperature change. Fig. 5 depicts the
potential range of aquifer change by 2080. The
strongest effect is produced by MIROC-A2 scenario,
which causes approximately 3.9 °C warming in
aquifer temperature at 8 m depth (the deepest depth
for the water table among five observation wells).
MRI-A2 scenario exhibits a moderate effect on
aquifer temperature change near to the ground
surface (2.8 °C at 8 m depth), but it is the robust in
terms of carrying the climate change impacts to the
deeper aquifer depths. As depicts in Fig. 5, by year
2080, depth of departure from steady state T-D
profile will approximately extend to 120 m. This is
because; background temperature in the aquifer is
higher than the annual average surface air
temperature, giving an upward heat flow from the
interior of the earth. Therefore, in the presence of
significant groundwater recharge, cool water
infiltration from the top of the aquifer eventually
slowdown the rate of aquifer temperature rise due to
surface air temperature change owing to climate
change. When MRI-A2 scenario produce substantial
groundwater recharge reduction (about 63% from
2007 estimation), it results to decelerate the cooling
process of infiltrating water flow and carry the
climate change signatures earlier than other
scenarios. When accounting the effect of the entire
model scenarios, aquifer temperature may change in
1.2-3.9 °C ranges which may have critical impact on
the ecological balance of the Sendai plain4).
between warming rate (WR) in future and bias of
GCM prediction compared to observed warming
rate was applied ({GCM WR2000-2099 – [GCM
WR1927-1999 – Observed WR1927-1999]}/ Observed
WR1927-1999). This index produces the highest value
for MIROC (3.2 for MIROC-A2), which clear
explains the highest impact prediction of
MIROC-A2 than the HADCM3 model in the local
scale. Similarly, even though HADCM3 produces
moderate change for precipitation at the GCM grid
scale (Fig.3b), HADCM3-A2 scenario gives the
highest index value (0.15) than the MIROC-A2
(0.1) or CCSM3 (0.06). This is again due to its low
bias of HADCM3 relative to the observed
precipitation than the MIROC or CCSM3 models.
As a result HADCM3-A2 makes the highest
prediction of future precipitation change from the
downscaled results at the local scale (345 mm/year).
(4) Potential climate change impacts at the local
scale aquifer thermal regimes
To account the climate change effects on
hydrology in the Sendai plain, we substituted the
projected precipitation and temperature to water
budget technique, which was previously developed
by Gunawardhana and Kazama1),2) for the Sendai
plain. The possible variations of groundwater
recharge were estimated with respect to the change
in surface runoff and evapotranspiration due to
changing climate in future. When comparing the
different GCMs performances, HADCM3 estimates
decreasing trend of groundwater recharge (1-26%
reduction from 2007 estimation) for its all scenarios.
As a result of its projected decreasing trend of
precipitation and moderate warming, MRI-A2
scenario predicts the strongest hydrological impact,
which reduces the groundwater recharge to 50
mm/year by year 2080 in the Sendai plain.
According to MIROC-A2, anticipated groundwater
recharge would decrease by 32 mm/year, despite the
1
1
a)
b)
Observed JMA
1967-2006 in Jan
0.8
Cumulative probability
Cumulative probability .
0.8
Downscaled 20602999 in Jan
0.6
Observed JMA
1967-2006 in Jul
0.4
Downscaled 20602999 in Jul
0.2
0.6
Observed JMA 1967-2006 in Aug
0.4
Downscaled 2060-2099 in Aug
Observed JMA 1967-2006 in Apr
0.2
Downscaled 2060-2099 in Apr
0
0
-5
0
5
10
15
20
25
30
35
0
Monthly temperature (oC)
100
200
300
400
Monthly precipitation (mm)
500
600
Fig. 4 Cumulative probability distributions of observations and MIROC-A2 predictions; a) temperature, and b) precipitation
- 71 -
Annual Journal of Hydraulic Engineering, JSCE, Vol.54, 2010, February
Temperature (°C)
0
13
14
15
16
17
-20
Depth (m)
-40
-60
-80
-100
-120
18
19
Observation in 2007
Geothermal gradient
Predictions by 2080
HadCM3 A2
HadCM3 A1B
HadCM3 B1
Mri A2
Mri A1B
Mri B1
ECHAM5 A2
ECHAM5 A1B
ECHAM5 B1
Miroc A2
Miroc A1B
Miroc B1
NCAR A2
NCAR A1B
NCAR B1
Fig. 5 Range of aquifer temperature change by different GCMs
4. CONCLUSIONS
GCMs output with various downscaling
techniques link the global scale climate features to
the local scale and make a platform for climate
change studies. However, different model structures
with various grid resolutions produce different result
for the same area of interest, which makes the use of
single model with few scenarios realistically critical.
We considered six GCMs with three scenarios in
each and examine their behaviors with respect to
temperature and precipitation in the grid box
containing the Sendai plain. Magnitude of the
climate parameter change in the future at the GCM
grid scale did not necessarily indicate the same
order of magnitude in downscaled results due to
existent model bias with observations at the local
scale. However, model bias may arise from various
effects, and may not necessarily indicate that the
model cannot correctly capture the large-scale
climate change signals. Therefore, two indexes were
introduced that accounts the combined effects of
model bias and the magnitude of climatic parameter
change in the future and we conclude that
HADCM3 and MIROC models may suitable for
climate change studies such as flood forecast due to
its moderate bias and higher precipitation arising
trend in future in the Sendai plain. Then again, MRI
model would suite for the quantity and quality
aspects of the water resources anticipated to climate
change. Selection of these three models (HADCM3,
MRI and MIROC) will simulate the potential
climate change impact in highest and lowest
extreme events, which may minimize the
computational cost of using several models.
The use of 18 scenarios from six GCMs in our
study generates a consistent range of impact on the
hydrology in the Sendai plain. With respect to
- 72 -
1.8-4.7 °C surface air temperature rise (minimum in
CSIRO-B1 and highest in MIROC-A2), 11 mm/year
precipitation reduction in MRI-A2, and 345
mm/year precipitation rise in HADCM3-A2,
groundwater recharge would vary in a range of
50-182 mm/year. When, different degrees of
groundwater recharge and ground surface warming
rates were incorporated, aquifer in the Sendai plain
may warm in 1.2-3.9 °C range by 2080.
According to IPCC AR4, approximately 20-30%
of animal species are likely to be at risk of
extinction if increases in global average temperature
exceed 1.5-2.5 °C. Therefore, estimated range of
aquifer warming will have a critical impact for the
ecological balance in the Sendai plain. Land-use
practices, such as reforestation have the potential to
decrease aquifer temperature by increasing shading.
Therefore, estimations in this study will be
important for mitigating climate change impacts,
where aquifer temperature increases due to ground
surface temperature change would likely
compensate by lowering aquifer temperature with
the effect of reforestation program.
ACKNOWLEDGMENT:
This
work
was
supported by the Global Environment Research
Fund (S-4) of the Ministry of Environment and
Grants-in-Aid for Scientific research, Japan. We are
also grateful to Mariko Izumi and Dr. Seiki
Kawagoe for supporting in observations.
REFERENCES
1) Gunawardhana, L. N. and Kazama, S.: Spatial downscaling
of GCM output for assessing the impacts on groundwater
temperature in the Sendai plain, Ann. J. Hydraulic. Eng., Vol.
53, pp.79-84, 2009a.
2) Gunawardhana, L. N. and Kazama, S.: Climate change
impacts on groundwater temperature change in the Sendai
plain, Japan, Hydrological Processes, Accepted, 2009b.
3) Oie, G. and Olsen, Y.: Influence of rapid changes in salinity
and temperature on the mobility of the rotifer Brachionus
plicatilis, Hydrobiologia, Vol. 255/256, pp.81-86, 1993.
4) IPCC (Intergovernmental Panel on Climate Change),
working group ΙΙ,: Impacts, Adaptation and Vulnerability.
Cambridge University Press, 2007.
5) Salathe Jr. E. P., Mote, P. W. and Wiley, M. W.: Review of
scenario selection and downscaling methods for the
assessment of climate change impacts on hydrology in the
United States Pacific Northwest, Int. J. Climatology, Vol.
27, pp.1611-1621, 2007.
6) Uchida, Y. and Hayashi T.: Effects of hydrogeological and
climate change on the subsurface thermal regime in the
Sendai Plain, Physics of the Earth and Planetary Interiors,
Vol.152, pp.292–304, 2005.
7) Carslaw, H. S. and Jaeger J.C.: Conduction of heat in solids,
2nd ed. Oxford university press, 1959.
(Received September 30, 2009)